Immersion fixation and staining of multi-cubic millimeter volumes for electron microscopy-based connectomics of human brain biopsies

Abstract

Connectomics allows mapping of cells and their circuits at the nanometer scale in volumes of about 1 mm³. Given that the human cerebral cortex can be 3 mm in thickness, larger volumes are required. Larger volume circuit reconstructions of human brain are limited by 1) the availability of fresh biopsies; 2) the need for excellent preservation of ultrastructure, including extracellular space; and 3) the requirement of uniform staining throughout the sample, among other technical challenges. Cerebral cortical samples from neurosurgical patients are available owing to lead placement for deep brain stimulation. Described here is an immersion fixation, heavy metal staining and tissue processing method that consistently provides excellent ultrastructure throughout human and rodent surgical brain samples of volumes 2 × 2 × 2 mm³ and of volumes up to 37 mm³ with one dimension ≤ 2mm. This method should allow synapse-level circuit analysis in samples from patients with psychiatric and neurologic disorders.

Introduction

The human brain is a vastly complicated tissue, and little is known about its microstructure, particularly the synaptic connectivity of its highly diverse and numerous neuronal cell types (1). These synaptic connections underlie our cognitive capabilities, and it is likely that their disruption gives rise to many presently incurable brain disorders. Recent advances in mapping nerve cell connectivity have come from connectomics via serial section electron microscopy (EM), which provides neuronal connectivity data at synaptic resolution (2–13). To study normal, and potentially disordered, human brain microstructure, cortical samples can be collected from patients with neurologic and psychiatric disorders undergoing neurosurgical interventions (14–17).

Although most non-human studies preserve tissue with trans-cardiac perfusion (2,4,5,11,18–20), immersion fixation is the only option for human biopsies generated in the clinical setting (21–27). Furthermore, in most immersion studies, the tissue volumes are much smaller than what would be required for full-thickness cortical neural circuits in humans, (≥ 3 × 1 × 1 mm³). In addition to excellent fixation, heavy metal staining needs to uniformly penetrate large surgical biopsies for EM. This article describes a protocol for immersion fixation, heavy metal staining and tissue processing that consistently provides excellent ultrastructure and maintenance of extracellular space (ECS) for full-thickness cerebral cortex biopsies (≥ 8 mm³).

Methods

Human brain biopsies were collected in collaboration with the Icahn School of Medicine at Mount Sinai, as approved by both the Institute Review Boards (IRB) at Mount Sinai and at Harvard University. Fourteen biopsies were collected from the sites of lead placement in the prefrontal cortex for deep brain stimulation. All brain biopsy samples were collected from patients who had provided informed consent for sample collection and public sharing of de-identified data. We also collected brain samples from adult C57BL/6 mice. All animal procedures were performed according to National Institutes of Health guidelines and approved by the Institutional Animal Care and Use Committee at Harvard University.

To obtain the human samples, the neurosurgeon drilled a burr hole in the frontal bone and made a pial incision to obtain a multi-cubic millimeter cortical biopsy using a bio punch via microdissection (L.E. Liharska, M.S., et al., unpublished data, 2023). The biopsy was placed on a dampened gauze pad and was removed from the sterile field. Using a metal probe to guide it, the biopsy was transferred into a vial containing 2 ml of cold fixative (see Table 1) and sealed. The time from biopsy extraction to immersion fixation was ≤ 60 seconds. The biopsy was later transferred to larger volume of chilled fixative and stored at 4°C until staining. Large biopsies were cut to smaller blocks ≤ 4 × 4 × 4 mm during the first 24 hours after extraction (Table 1, Figure 1A).

Table 1.

Method for collection, immersion fixation and heavy metal staining of 2 × 2 × 2 mm³ brain tissue blocks.

Step #	Process	Time	Temperature
Collection and Immersion fixation of human and mouse biopsies
1a.	Human: Rapidly transfer the excised neurosurgical biopsy to a small vial containing 2–4 ml of cold fixative.	1–5 min	4°C
	Transfer biopsy to larger volume (50–100 ml) of cold fixative.	within 5 min of excision
	Cut biopsies into volumes no larger than 4 × 4 × 4 mm3 and return to cold fixative.	≤24 h after excision
	Store in refrigerator for days to weeks depending on biopsy volume.	3–28 days
	Fixative: 2.5% GA, 2% PFA, 2mM CaCl2, 0.15M NaC buffer with 3% Mannitol (445 mOsm), pH 7.4.
	Prepare fresh fixative within 1–2 h of use and keep it at 4°C.
1b.	Mouse: Anesthetize an adult mouse with isoflurane, euthanize by cervical dislocation and remove the brain.	1–2 min	RT
	Place brain in petri dish containing cold fixative on ice and punch or excise 2 × 2 × 2 mm3 tissue blocks.	1–3 min	4°C
	Transfer these blocks to 20 ml vials containing cold fixative and store in refrigerator.	72 h
	Fixative: 2.5% GA, 2% PFA, 2mM CaCl2, 0.15M NaC buffer with 3% Mannitol (445 mOsm), pH 7.4.
	Prepare fresh fixative within 1–2 h of use and keep it at 4°C.
Heavy metal processing of 2 × 2 × 2 mm3 blocks or blocks with smallest dimension = 2 mm
2.	Buffer wash:0.15M NaC (Ph 7.4) + 2mM CaCl2	1 × 30 min	4°C
		2 × 30 min	RT
3.	2% OsO4 in 0.15M NaC (Ph 7.4) + 2mM CaCl2	3 h	RT
	Replace with fresh 2% OsO4 in 0.15M NaC (Ph 7.4) + 2mM CaCl2	8–12 h	4°C
4.	Buffer wash:0.15M NaC (Ph 7.4) + 2mM CaCl2	3 × 15 min	RT
5.	2.5% Potassium ferrocyanide, in 0.15M NC (Ph 7.4) + 2mM CaCl2	3.3 h	RT
6.	Buffer wash:0.15M NaC (Ph 7.4) + 2mM CaCl2	3 × 30 min	RT
7.	Filtered, 0.8 % (w/v) TCH in Normal saline + 1.2% mannitol	45 min	RT
8.	Distilled water wash following double rinse with NS	1 × 40 min	RT
		2 × 30 min	RT
9.	2% Aqueous OsO4	5 h	4°C
10.	Distilled water wash	3 × 30 min	4°C
11.	2% Aqueous uranyl acetate	8–12 h	4°C
		2 h	RT
12.	Distilled water wash	3 × 45 min	RT
13.	Dehydration with graded acetonitrile in distilled water		RT
	10% Acetonitrile / 90% distilled water	20 min
	25% Acetonitrile / 75% distilled water	20 min
	50% Acetonitrile / 50% distilled water	20 min
	75%Acetonitrile / 25% distilled water	15 min
	100% Acetonitrile	2 × 15 min
14.	Infiltration with series dilution of Resin in acetonitrile		RT
	25% resin / 75% acetonitrile	4 h
	50% resin / 50% acetonitrile	12 h
	75% resin / 25% acetonitrile	12–18 h
	100% resin (includes one exchange with fresh resin)	24 h
	Resin: LX-112 (31.28 g), NMA (19.4 g), NSA (9.4g) and BDMA (1.2g). Critical: 1. Prepare fresh resin for 75% and 100% exchange
15.	Embedding in 100% LX-112 resin (recipe as in step 14)	-	RT
16.	Curing	4 h	45°C
		68 h	60°C
Critical: 1. All solutions (Step 2–15) are prepared fresh.    2. Incubation at each step is done in 15–20 ml solutions.
BDMA, N,N dimethyl benzylamine; GA, glutaraldehyde; NaC, sodium cacodylate; NMA, methyl 5-norbornene-2,3-dicarboxylic anhydride; NSA, nonenyl succinic anhydride modified; NS, normal saline; PFA, paraformaldehyde; RT, room temperature; TCH, thiocarbohydrazide. See key resource table for reagent details.

Description of steps for collection, immersion fixation and heavy metal processing of human and mouse brain biopsy samples of volume 2 × 2 × 2 (8) mm³ and of volume > 8mm³ with at least one dimension ≤ 2mm.

Figure 1: Immersion fixation and heavy metal staining of human and mouse cortical biopsies.

A, Shown, are the steps for immersion fixation and heavy metal staining of large brain biopsy tissue blocks. B, Electron micrograph from a human brain biopsy (7.25 × 3 × 1.2 mm³) taken at a z depth of ~300 μm. The arrows point to synapses, the arrowheads show spine apparatus, “M” are mitochondria, and the asterisk shows myelin C, Electron micrograph from mouse cortical block (2 × 2 × 2 mm³) taken at a z depth of 1mm. The arrows, etc. are as in panel B. D, Electron micrograph from the same human cortical biopsy as shown in panel (B). The ECS is shown in red and is 14.5%. E, Electron micrograph from a different human cortical biopsy (8.25 × 2.5 × 1.8 mm³) at center of x-y and a z depth of ~1mm with 12.2% ECS highlighted in red. F, Electron micrograph of cortex from immersion fixed mouse biopsy (shown in C) with 13.4% ECS shown in red. G, Electron micrograph from mouse biopsy (2 × 2 × 2 mm³) that was obtained after trans-cardiac perfusion. ECS is diminished (4.9%). H, Electron micrograph from immersion fixed mouse cortex showing a spinule at a cortical synapse (arrow). I, Electron micrograph from human cortex showing an omega figure at a synapse (arrow). J, Graph showing a significant increase in omega figures in immersion fixed versus trans-cardiac perfused brain samples (*** p<0.0001, Chi-square test; error bars show SEM). CaCl₂, calcium chloride; ECS, extracellular space; GA, glutaraldehyde; IF, immersion fixed; NaC, sodium cacodylate; PF, perfusion fixed; PFA, paraformaldehyde; TCH, thiocarbohydrazide.

To obtain rodent samples, mice were anesthetized with isoflurane (Baxter) inhalation and euthanized by cervical dislocation. Within 90 seconds of euthanasia, the cerebral cortex was exposed and the brain rostral to the spinal cord was removed and immersed in cold (4°C) fixative (Table 1, Figure 1A). To match the clinical retrieval procedure, we used punch biopsies (Ted Pella, Inc. USA, diameter 1–4 mm). Four to ten biopsies (along the dorsal-ventral axis) were collected from each mouse brain and were transferred to vials containing cold fixative by within 5 minutes from the time of euthanasia (Table 1, Figure 1A). For trans-cardiac perfusion fixation, mice were anesthetized with isoflurane and perfusion was performed as described earlier (9). We used cold fixative (Table 1) for perfusion and then followed same steps as for the immersion fixation, described above.

The method for immersion fixation, heavy metal staining, resin infiltration and embedding of samples is described in Table 1 (see Key Resource Table for specific reagents used). Screening of samples, sectioning, and image acquisition were done as performed previously (15,28–32), and described in Supplementary Methods (also see Figure 2C, 2N). Image processing, stitching, alignment, and segmentation of serial section imaged datasets is described in Supplementary Methods (also see Figure 2).

KEY RESOURCES TABLE.

Resource Type	Specific Reagent or Resource	Source or Reference	Identifiers	Additional Information
Add additional rows as needed for each resource type	Include species and sex when applicable.	Include name of manufacturer, company, repository, individual, or research lab. Include PMID or DOI for references; use “this paper” if new.	Include catalog numbers, stock numbers, database IDs or accession numbers, and/or RRIDs. RRIDs are highly encouraged; search for RRIDs at https://scicrunch.org/resources.	Include any additional information or notes if necessary.
Chemical Compound	Glutaraldehyde 25% Aqueous Solution	Electron Microscopy Sciences	#16220
Chemical Compound	Paraformaldehyde 16% Aqueous Solution	Electron Microscopy Sciences	#15710
Chemical Compound	Calcium Chloride 1M	Sigma-Aldrich	#21115
Chemical Compound	D-Mannitol	Sigma-Aldrich	#M4125
Chemical Compound	Lactated Ringer's Solution	B Braun Medical Inc	#BMGL7500
Chemical Compound	Cacodylic Acid- Sodium Salt, TrihydrateSodium Cacodylate	Electron Microscopy Sciences	#12310
Chemical Compound	4% aqueous Osmium Tetroxide	Electron Microscopy Sciences	#19190
Chemical Compound	2% aqueous Osmium Tetroxide	Electron Microscopy Sciences	#19173
Chemical Compound	Potassium Ferrocyanide	Electron Microscopy Sciences	#26604–01
Chemical Compound	TCH (Thiocarbohydrazide)	Electron Microscopy Sciences	#21900
Chemical Compound	Sodium Chloride solution 0.9%	Sigma-Aldrich	#S8776
Chemical Compound	2% Uranyl Acetate	Electron Microscopy Sciences	#22400
Chemical Compound	Acetonitrile	Electron Microscopy Sciences	#10020
Chemical Compound	Distilled water (Ultra Pure)	Invitrogen	#10977023
Chemical Compound	Nonenyl Succinic Anhydride,	Electron Microscopy Sciences	#19050
Chemical Compound	NMA, (Methyl-5-Norbornene-2,3-Dicarboxylic Anhydride)	Electron Microscopy Sciences	#19000
Chemical Compound	LX 112 Kit with BDMA	LADD Research Industries	#21212
Chemical Compound	Sodium Gluconate	Sigma-Aldrich	#S2054
Chemical Compound	CFTR Inhibitor-II, GlyH-101	Sigma-Aldrich	#219691
Chemical Compound	DIDS	Sigma-Aldrich	#D3514

Figure 2: Serial section volume collection, imaging, alignment and segmentation from human and mouse cortical biopsy (Bx) blocks.

A, An immersion fixed human brain biopsy (7.25 × 3 × 1.2 mm³). B, The biopsy shown in A, processed with the heavy metal staining (2 × 2 × 2 mm³ protocol). The red and blue arrow heads in A and B show the same regions. Dotted white lines show sites of trimming before resin embedding. C, An X-ray tomogram of the same stained sample, z-depth ~300 μm. Dotted white lines as in B. The blue box highlights the region from where a series of 1175, 30nm sections were collected. D, A low-resolution electron micrograph (3 × 2.5 mm²) taken from region highlighted in blue box in C. Blue arrowheads are same as in A-C. The black rectangle shows the region of volume imaged with multibeam scanning electron microscope. E, Low resolution 750 × 250 μm² electron micrograph of highlighted rectangle in D. F, An aligned stack of 160 × 160 × 35 μm³ from the region highlighted with yellow box in E. G, An electron micrograph with cluster of axons traced manually from volume in F. H, Rendering of the manually segmented axons from G; all of them could be traced to two edges of the volume. I, Rendering of a spiny dendrite traced from volume in F. J, Inset region from I, (changed orientation) showing presynaptic axons making synapses on spines of the blue-colored dendrite in I. K, Electron micrograph of region in white square in J, showing a synapse from the red-colored axon onto a dendritic spine. L, Rendering of a manually segmented astrocytic process traced to astrocyte soma. M, A magnified view of an astrocytic process (yellow) and glycogen granules (magenta) from area within white box in L. Arrow in M points to a thin process of astrocyte. N, A low-resolution electron micrograph (shown with arrow) superimposed on x-ray tomogram (shown with asterisk) at z-depth of 1 mm from a 2 × 2 × 2 mm³ mouse brain biopsy. The yellow box shows 300 × 300 μm² region from where series of ~100, 30 nm sections were aligned and automatically segmented. Automatically segmented region from the blue box is shown in O. P, An electron micrograph with automatic 2D segmentation shown in Q. R, Rendering of two automatically segmented spiny dendrites (green and yellow as seen in Q). S, Cluster of automatically segmented axons from yellow boxed area in N. T, Three-dimensional rendering of the axons shown in S. U, A rendering of the 8000 largest objects segmented automatically from region highlighted with yellow box in N.

Results

Immersion Fixation

Based on the assessment of many fixation variables (Supplementary Figure 1) we found that 2.5% glutaraldehyde, 2% paraformaldehyde and 2mM calcium chloride (CaCl₂) in 0.15M sodium cacodylate (NaC) buffered to pH 7.4 provided excellent fixation with immersion in both human and mouse samples. Using 0.15M NaC (295 mOsm) rather than 0.1M NaC (200 mOsm) helped to maintain the ECS as previously noted (33,34) but was not effective in large volume samples with a depth of >1mm. We therefore tested different concentrations of the sugar, mannitol, known to reduce edema and control intracranial volume in patients with traumatic brain injury (35). Mannitol helped preserve ECS when used at 2% (1 × 1 × 1 mm³), 3% (2 × 2 × 2 mm³), and 3 or 3.5% (3 × 3 × 3 mm³) in the fixative solution. However, a higher mannitol concentration (4% w/v) resulted in severe shrinkage of fine cell processes (36,37).

Other ECS maintenance strategies were also tested including sodium gluconate and the anion channel regulators—DIDS (4,4’-diisothiocyanatostilbene-2,2’-disulfonic acid) and the cystic fibrosis transmembrane conductance regulator antagonist glycine hydrazide (GlyH-101)—as these are thought to reduce neuronal swelling (38). These additional strategies to maintain the ECS were ineffective. Interestingly, DIDS and sodium gluconate noticeably improved contrast at sites of synaptic contact (Supplementary Figure 2, Key resource table).

The temperature of the fixative during immersion, and the time from tissue harvesting until immersion had strong effects on the ultimate quality of the ultrastructure. A comparison of fixation of biopsy samples based on temperature was undertaken to assess chilled fixative (4°C) versus room temperature (RT) fixative. Biopsy samples preserved directly in cold fixatives always had better overall quality when compared to samples immersed in fixative at RT as previously documented (39). In testing for an acceptable time delay between the tissue harvesting and immersion fixation, we found that an interval of ≥ 10-minute between immersion in fixative-free lactated Ringer’s (270 mOsm; see Key resource table) and the fixative was unacceptable. In particular, ≥ 10-minute delays caused three ultrastructural abnormalities: broken membranes, mitochondria whose cristae were poorly defined, and large cytoplasm-free vacuoles. Therefore, samples were immersed in cold fixative within 1 minute (human) and within 5 minutes (mouse) after tissue harvesting.

There was little prior information on the duration of immersion fixation for good ultrastructure. To determine the minimal duration of the fixation required an analysis of immersion fixation was undertaken for blocks 1 × 1 × 1 mm³ and 2 × 2 × 2 mm³. At 30 hours for 1 × 1 ×1 mm³ samples and 46 hours for 2 × 2 × 2 mm³ samples, the center of the volumes had lower contrast than the periphery as determined by Scanning electron microscopy. In addition to less intensity, the middle of the sample showed poor membrane definition, poor tissue integrity (many small sites where the tissue pulled apart), poor definition of mitochondrial cristae, and difficulty recognizing synapses. Low staining contrast is often ascribed to poor penetration of osmium tetroxide, however, in this case the problem was explained by incomplete fixation because the same osmium staining protocol resulted in well-preserved ultrastructure when the fixation step was prolonged to 46 hours (1 × 1 × 1 mm³) or 72 hours (2 × 2 × 2 mm³). For 3 × 3 × 3 mm³ and 4 × 4 × 4 mm³ blocks, fixation for 7–28 days gave good ultrastructure. Finally, better membrane preservation of cells and cellular organelles occurred when fixative contained 2mM CaCl₂ versus its absence (40,41).

Heavy Metal Staining

As described in detail in the flowchart (Supplementary Figure 1), many staining parameters were altered to achieve uniformly strong osmium and uranium staining without cracks or brittleness in immersion-fixed samples that were at least 2 × 2 × 2 (8) mm³. The details of each step are provided in Table 1. Eight critical aspects of this protocol are:

1. Multiple NaC buffer washes after aldehyde fixation to lower osmium’s opportunity to react with aldehydes, to improve middle-of-the-block staining (42)

2. Osmium tetroxide incubation divided into first, a 3h incubation at RT to fix the lipids and second, a 12h step at 4°C for staining

3. An NaC buffer wash prior to potassium ferrocyanide reduction to remove unbound osmium

4. Decreased duration of potassium ferrocyanide to a time less than the first osmium step to mitigate expansion, brittleness of the tissue

5. Addition of 2mM CaCl₂ at all steps prior to thiocarbohydrazide (TCH) to increase membrane contrast and lessen breaks

6. Increasing the osmolarity of the vehicle plus reducing the concentration, temperature, and time of TCH reaction to prevent cracks

7. Decreased duration of the second OsO₄ staining to a time less than the first osmium step and lowering the incubation temperature to improve contrast

8. Incubation with lower temperature uranyl acetate to reduce cytoplasmic darkening

Based on these modifications, the tissue fixation and the staining results were excellent in 2 × 2 × 2 (8) mm³ blocks and in blocks > 8mm³ when at least one dimension of the block was ≤ 2 mm (Figure 1 B–F; Figure 2 C, D, N and Supplementary Figures 3–5). To investigate the possibility that this protocol might also be useful for blocks where all three dimensions were > 2 mm, incubation times of the various steps were increased, but this approach resulted in poor staining in the core and cracks. For 3 × 3 × 3 (27) mm³ blocks, more uniform heavy metal staining without cracks was accomplished once (after assessing a range of protocols in 21 blocks; see Supplementary Figure 1), using a modified protocol (Supplementary Table 1) that showed sufficient staining at the center point, 1.5 mm from the nearest surface (Supplementary Figure 6).

To investigate whether this high quality of the ultrastructure could also be obtained with perfusion fixed, as opposed to immersion fixed, brain samples, punch biopsies were removed from trans-cardiac perfusion fixed mouse brains samples (see Methods), followed by the staining protocol described in Table 1. The ultrastructure was similar in appearance and quality to that obtained with immersion fixation (Supplementary Figure 7), although the ECS was reduced (Figure 1G). In a 2 × 2 × 2 mm³ immersion fixed mouse sample, ECS was on average 15% (20% in Layer 1 and 13% in Layer 6, SEM ± 1.3). In the human immersion fixed samples ECS was on average 16% (25% at surface and 12% at center of block, SEM ± 2.8) (see Figure 1D–F). In contrast in a 2 × 2 × 2 (8) mm³ transcardially perfused sample, the mean ECS was <5% (range 4.9–4.2, surface to center, SEM ± 0.35, Fig 1G; also see Supplementary Methods).

Ultrastructure in immersion fixed samples

Use of immersion fixed samples for connectomic studies requires that synapses, glia, myelin, blood vessels and fine structure is well preserved. The immersion-fixed samples showed well stained synapses, myelin, mitochondria, blood vessels and other ultrastructural features that often show abnormalities if poorly fixed or stained (43,44) (Figure 1B–C, Figure 2M and Supplementary Figure 4) also see the EM data at (https://lichtman.rc.fas.harvard.edu/mouse_cortex_at_1mm); Supplementary Figure 4). Even in the center of a block that was 8.25 × 2.5 × 1.8 (37) mm³, the ultrastructure was well preserved (Supplementary Figure 3). Serial-section stacks of electron micrographs from these blocks revealed several unusual features at sites of close contact between cells, some of which have been reported previously (45–50). For example, small dendritic spinules inserted into axonal boutons (Figure 1H; Supplementary Figure 8C) and sometimes whole spine heads invaginated into presynaptic axons (Supplementary Figure 8E–F). Axo-axonal invaginations (Supplementary Figure 8D), axo-astrocytic and dendro-astrocytic invaginations near synapses (Supplementary Figure 8G–H), endothelial-endothelial cell invaginations and pericyte invaginations into the cytoplasm of endothelial cells (Supplementary Figure 8I–J) were all commonly observed. Red blood cell interactions with endothelial processes in blood vessels were also common. Endothelial microvilli invaginating red blood cells (51) and what appeared to be ejected membrane-bound material from red blood cells engulfed by endothelial cells were also observed (Supplementary Figure 9).

Occasionally what appeared to be fused vesicles at synapses with an omega shape were observed (Figure 1I), (52–56). These omega figures were seen in both mouse and human brain tissue. About 65% of the synapses in the aligned human cortical sample showed omega figures (64% in excitatory synapses and 70% in inhibitory synapses; Supplementary Figure 8A, Supplementary methods). Similarly, 59% of the synapses in the aligned mouse cortical sample had omega figures (57% in excitatory synapses and 62% in inhibitory synapses; Supplementary Figure 8B). Given the rarity of seeing omega figures in previous electron microscopy datasets, it was possible that these profiles were related to immersion fixation. Indeed, the incidence of seeing omega profiles was 8-fold greater in immersion fixed than in perfused tissue (p<0.0001, Chi-square test; see Figure 1J and Supplementary Methods.

Membrane breaks

Given the deleterious effect of membrane breaks on automatic segmentation, they were assessed for each membrane bound profile in serial-section stacks from samples of immersion fixed mouse and human blocks that were at least 2 × 2 × 2 (8) mm³ (Supplementary Methods). There were few breaks: 0.05% in mouse and 0.07% in the human data.

Automated and manual segmentation

The ultimate appraisal of ultrastructure quality in connectomics rests on whether neural circuits can be reconstructed. To evaluate this, 1175 images from 30 nm sections were collected at ~300 μm depth from a human sample (Figure 2A–F; see Supplementary Methods). Manual reconstruction was straightforward. Eight axons were traced identically by three independent annotators. All these axons could be traced to at least two edges of the volume (Figure 2G–H). In addition, two challenging aspects of connectomics data were also traced without difficulty: the spines originating from a spiny dendrite and the many fine terminal processes of an astrocyte (Figure 2I–M). Traced astrocytic processes sometimes narrowed to diameters of less than 40 nm (Figure 2M), nonetheless they could be followed until turning to larger diameter process or leaving the volume. All these results argue that these large immersion fixed volumes had excellent staining. Automatic segmentation on a 300 × 300 × 3μm³ aligned volume from the center (1 mm depth) of an 2 × 2 × 2 mm³ mouse brain sample was also tested, (Figure 2N–U, see segmentation layer at https://lichtman.rc.fas.harvard.edu/mouse_cortex_at_1mm; Supplementary Methods). Figure 2U shows the largest 8000 segmented objects and 77% of these touched the edges of the volume. The negligible membrane breaks in this dataset (mentioned above), mitigated most in-plane merge errors. The few merge errors observed were secondary to substrate artifacts as opposed to broken or too faint membrane staining, also arguing for the high quality of the immersion fixation and subsequent staining.

Discussion

The goal of this project was to find a method that provided excellent fixation and ultrastructural staining for fresh neurosurgical human biopsies where perfusion with fixatives was not possible. Because one use of this tissue is connectomics, it was also important that the approach worked well with volumes much greater than 1 mm³ and had a sufficient amount of ECS to make tracing less difficult. Generating this protocol required modifications to many fixation and heavy metal staining parameters. Each of the staining and washing steps is fraught with some uncertainty about the underlying mechanisms and optimizing osmium staining is further complicated because this heavy metal has multiple valence states with different reactivities (57). An empirical approach was used to compare the quality of fixation, staining and tissue penetration using 147 protocol variables in 237 samples and ultimately home in on an optimal protocol (see Supplementary Figure 1).

Several principles of immersion fixation and staining emerged: First) Fast transfer (≤5 min) of fresh biopsies to fixative was important. Second) Cool temperature (4°C or RT) was helpful in all steps except resin curing; Third) ECS preservation using mannitol aided fixation, osmium penetration and segmentation, see also (34); Fourth) Extensive washing to clear fixatives and unbound osmium gave good stain contrast between membranes and cytoplasm; Fifth) A longer first osmium step and a shorter reducing step improved staining and tissue integrity; Sixth) The addition of CaCl₂ reduced membrane breaks. In contrast to previous studies, this protocol uses CaCl₂ to stabilize membranes from the initial aldehyde fixation step through the potassium ferrocyanide step (4,40,41,45). Seventh) The TCH step was improved by using a lower concentration, and in saline as opposed to water. This approach reduced osmotic swelling and gas formation; Eight) The second osmication needs far less time than the first.

The rapid cold fixation with maintenance of ECS gave ultrastructural features that are less commonly found in traditional cardiac perfused samples. For example, dendritic spinules and larger invaginations of dendrites into the presynaptic terminal were seen. Such features have been previously described but in smaller blocks and in some samples pristinely prepared via high pressure freezing (45–50,58,59). Additionally, glycogen granules were observed in astrocytic processes as have been previously described (60–64). However preservation of these granules is often challenging (62,65) suggesting that the rapid cold immersion approach preserves labile features well. Invaginating processes of blood vessels cells were also seen. Pericytes invaginated endothelial cells and endothelial cells invaginated each other in ways that were distinct from the classic peg and socket (66) and intercellular tight junctions (67). Finally, many synaptic sites showed synaptic vesicles that were fused with presynaptic terminal membrane (omega figures). Perhaps, these were explained by the rapid cold fixation freezing synapse release in the act. We cannot however discount the possibility that hypertonic mannitol fixative induced vesicle fusion because hypertonic sugar solutions have been used to enhance vesicle fusion events (54,68). One additional advantage of immersion fixation was the more normal ultrastructure of cerebral microvasculature. A caveat of studying microvasculature with EM is that perfusion removes blood cells limiting analysis of interactions between blood cells and the blood vessel wall. Numerous processes extending between vessel wall and blood cells were observed in these samples. It is hoped that high quality preservation of human brain ultrastructure will ultimately provide insight into neurological and psychiatric questions that have been less accessible with previous approaches.